Dimension inconsistency in convolusion operations in Conv2d and F.conv2d

[alexzhuuuu]

Run the command like this:
"python tools/test.py configs/nuscenes/det/transfusion/secfpn/camera+lidar/swint_v0p075/convfuser.yaml pretrained/bevfusion-det.pth --eval bbox"

and I got an error like the following sentence:
'RuntimeError: Given groups=1, weight of size [8, 1, 1, 1], expected input[6, 6, 256, 704] to have 1 channels, but got 6 channels instead'

Problem of nn.Conv2d and F.conv2d??
How can I fix it?

[alexzhuuuu]

similar with #478
https://github.com/mit-han-lab/bevfusion/issues/478
Could anyone give some explainable solutions?